financial market failures and systemic riskfinancial market failures and systemic risk 3 resume dans...

No 1996 – 01January

Financial Market Failures and Systemic Risk

_____________

Michel Aglietta

CEPII, Document de travail n° 96-01

2

TABLE OF CONTENTS

RESUME .................................................................................................................................3

SUMMARY..............................................................................................................................4

INTRODUCTION......................................................................................................................5

I. RECENT TURMOILS IN FINANCIAL MARKETS...................................................6

I.1. FOREIGN EXCHANGE MARKETS : EMS AND DOLLAR CRISES............................................6I.2. THE 1994 WORLDWIDE BOND MARKET SLUMP..................................................................8I.3. FUTURES MARKETS : THE DEMISE OF BARINGS AND METALLGESELLSCHAFT....................9I.4. EMERGING MARKETS : THE MEXICAN CRISIS...................................................................13

II. DERIVATIVES AND FINANCIAL FRAGILITY.....................................................15

II.1. LESSONS FROM THE EPISODES : THE SOURCES OF SYSTEMIC RISK IN FINANCIAL MARKETS. ................................................................................................15

i. Market illiquidity can force banks acting as market makers to rely on dynamic hedging and to effectively convey the liquidity gap onto the underlying markets..............................................................................................16ii. Numerous reasons account for the possibility of one-way selling in present-day markets where competition is intense and price expectations are affected by complex parameters........................................................................17

II.2. DESTABILIZING PRICE DYNAMICS IN PERIOD OF STRESS: THE ROLE OF OPTIONS.............17II.3. UNCERTAINTY ABOUT CREDIT RISK: THE SWAP MARKET. ..............................................19II.4. CONCENTRATION OF MARKET MAKING IN OTC DERIVATIVES........................................21

CONCLUSION........................................................................................................................23

REFERENCES........................................................................................................................27

LIST OF WORKING PAPERS RELEASED BY CEPII ...............................................................29


3

RESUME

Dans les dernières années une succession d'accidents financiers s'est produite . Cesépisodes ont concerné une grande variété de marchés financiers incluant des marchésdérivés. Il y a un grand pas entre des perturbations locales et des crises financières étendues.Il est néanmoins utile d'étudier des évènements récents qui ont fait apparaitre desaugmentations importantes de volatilité et des problèmes de liquidité. Cette revue estconduite de manière à mettre en évidence le potentiel de propagation des perturbations quicrée un risque systémique latent.

Des conclusions générales peuvent être tirées sur les sources du risque systémiquedans les marchés financiers. Trois types de perturbations ont la possibilité de se propagerentre les marchés : les dynamiques de prix déstabilisantes lorsqu'elles se produisent enpériode de tensions macroéconomiques; l'incertitude pour évaluer le risque de crédit dansdes marchés de gré à gré où règne une asymétrie d'information très forte; les manquesbrutaux de liquidité de liquidité dans les marchés peu profonds. Ces trois sources deperturbations sont susceptibles de provoquer des discontinuités dans les engagementsfinanciers.

C'est ainsi que l'illiquidité de marché peut forcer les banques qui en sont les teneurs àprocéder à des couvertures dynamiques qui véhiculent le besoin de liquidité non satisfaitd'un marché à un autre sous l'effet d'une cascade de ventes à sens unique. De nombreusesraisons, révélées par les accidents financiers étudiés, rendent compte de la possibilité d'uncourant de ventes à sens unique dans les marchés actuels où la concurrence est intense et oùl'influence réciproque des opérateurs est grande.

Les marchés dérivés de gré à gré sont des sources importantes de risque pour lesinstitutions financières qui les émettent, les teneurs de marché sont peu nombreux etconcentrés. Ces marchés peuvent donc manquer de profondeur en période de forteincertitude sur les conditions macroéconomiques. Si ces conditions conduisent lesutilisateurs finaux de ces marchés à des ventes massives, des problèmes de liquidité peuventse manifester. La vulnérabilité de ces dysfonctionnements locaux au risque de propagationdépend de paramètres caractéristiques des processus stochastiques gouvernant leschangements de prix et d'éléments des structures de marché que la présente étude s'efforced'identifier.

La transformation du risque de marché en risque de sytème n'est toutefois pas sansremède. Plusieurs améliorations sont envisagées par les superviseurs : l'utilisation desmodèles internes de mesure des risques pour remodeler le contrôle prudentiel, l'énoncé derecommandations contraignantes pour une divulgation plus transparente des risque demarché et de crédit, l'élaboration et la mise en oeuvre de simulations des conséquences durisque systémique, l'application de provisions en capital plus exigeantes que les risques demarché.


4

SUMMARY

In the last few years, a host of disturbances arose in a whole range of financial andderivative markets. I review a few episodes which exhibited huge increases of volatility andliquidity problems and which had a potential for spreading over, thus entailing systemicrisk.

Common lessons can be drawn from the episodes as far as sources of systemic riskare concerned. Three main factors conducive to potential spillover effects between marketsstand out : destabilising price dynamics under conditions of stress, uncertainty about creditrisk assessment, vulnerability to market liquidity shortfalls.

Market illiquidity can force banks acting as market makers to rely on dynamichedging and to effectively convey the liquidity gap from one market to another under thepressure of one-way selling. Numerous reasons account for the possibility of one-wayselling in present day markets where competition is intense and price expectations areaffected by complex parameters.

OTC derivatives can act as weak links in periods of large unexpected changes inthe volatility of financial prices. Options which are highly sensitive to the volatility of theunderlying markets exacerbate price changes under stress. Potential future credit exposureon swap portfolios is very difficult to measure and can change markedly and abruptly withchanges in interest rates and exchange rates.

Because OTC derivatives are major sources of risk for market-making firms, thosemarkets often lack depth. Fixed investment and learning costs being substantial, marketmakers are concentrated ; which can lead to liquidity problems under the heavy selling ofend-users motivated by a shift in market sentiment. The vulnerability to spillover effectsdepends on parameters of stochastic price dynamics and of market structure which arehighlighted in the paper.

The exposure to market risk is not without remedy. Major issues include the use ofinternal models of risk measurement to frame prudential regulation, the guidelines for theappropriate disclosure of market and credit risks characteristics, the development of stresstesting, the requirement of more demanding standards including capital charges on marketrisks.


5


Michel Aglietta1

INTRODUCTION

In the last few years episodes of disturbances in financial markets have occurredfrequently. The present paper is reviewing the most spectacular ones. The purpose is not toprovide a detailed account of each episode but to find out what is common in the unstabledynamics of various markets.

The episodes under review are dissimilar by the magnitude of the lossesencountered and by the seriousness of the spillover effects into the financial system. Butthey have common characters : they are international in scope even when they occur in aparticular country. They show concern for disturbing price volatility and liquidity problems: they have a potential for spreading over and entailing system risk.

After drawing lessons from recent events, the concern for systemic risk is revisited.Monetarists often argue that financial market crises are pseudo-crises because they do nothave the potential to induce a global contraction of liquidity in the banking system as bankruns do [Schwartz, 1992]. On the opposite side economic historians have a broad view offinancial crises [Kindleberger, 1978]. But their theoretical foundations have been criticisedfor being too shaky or too eclectic [Mishkin, 1991]. More recently, however, it was shownthat crises bursting into financial markets may exhibit liquidity problems of the sortencountered in bank runs [Davis, 1995].

The present analysis will follow suit. It will argue that liquidity is at the gist ofsystemic risk. A unified theory can be held based upon liquidity problems. But the wayliquidity is generated in contemporaneous unregulated financial systems makes financialmarkets vulnerable to the externalities that feed on the dynamics of systemic risk.

In the second part of the paper, the role of derivatives is investigated in relationwith their involvement in recent events and with their contribution to liquidity-generatingprocesses in financial markets . Liquidity problems can occur as the side effect ofuncertainty in valuing credit risks in OTC markets, as the outcome of destabilising pricedynamics in periods of stress, as the consequence of a too concentrated marketmaking infragile segments of derivatives markets.

A conclusion outlines the prudential issues related to the problems highlighted in thepaper and the solutions framed into the recent G10-BIS proposals.

1 Scientific Advisor for the CEPII


6

I. RECENT TURMOILS IN FINANCIAL MARKETS

In recent years a large range of markets were hit by unexpected price volatilitytriggering destabilising speculation and bringing heavy losses. The foreign exchangemarkets experienced the EMS and the dollar crises. The worldwide slump in the bondmarkets in early 1994 came as a total surprise. The futures markets precipitated the de factocollapse of Barings and Metallgesellschaft, two venerable and respected companies. Andthe appealing emerging markets were badly shaken by the Mexican crisis.

In all those episodes I will emphasize the linkages which seed the vulnerability tosystemic risk. In most cases, methods for hedging risk with the use of derivatives areamong the self-reinforcing linkages.

I.1. Foreign Exchange Markets : EMS and dollar crises

The EMS crisis of September 1992 exhibited speculative attacks of anunprecedented magnitude. For the first time, these attacks occurred without the dollarplaying a major role. For the first time also, hedging strategies made an extensive use ofcurrency options.

The Lira was an early example of the new mood among currency traders. In theearly 90's, after the Italian government had lifted capital controls and decided to join thenarrow-band ERM, international investors were attracted by a high-inflation, high-yieldingcountry, promising future convergence. Under the rationale of the so-called convergencetrade, unusually large capital inflows poured unhedged into Italian financial markets, as theydid in Spain too. The open positions in Lira and Peseta reflected the confidence either thatthe currencies would not be devaluated before investments were maturing, or that the debtmarkets would be liquid enough to get off with limited losses which would not cancel outthe high income yield.

However, as the aftermath of the Maastricht Treaty went wrong, internationalinvestors wanted to lock in the interest yield while getting rid of the currency risk. Theybought put options with a strike price a the upper limit of the band in the EMS. Lirasecurities-cum-put options on the Lira made compound investments which severed currencyand interest rate risks. As long as the band was credible, the options were out-of-the money,thus worthless. But after the setback of the Danish referendum, pressures began to build upon intrinsically weak currencies in the EMS. When the exchange rate regime became nolonger credible, the interest rate on the Lira increased with the gap between the spotexchange rate and the central parity. The closer the spot rate to the upper limit of the band,the larger the spread between Italian and German short-run interest rates. Because thespread is equal to the forward discount on the Lira, a time came when the forward Lira-DMexchange rate exceeded the limit of the band. The put options were in-the-money and dulyexercised. The banks which had written the options beforehand resorted to dynamichedging. They rushed to sell Liras heavily on the spot market against the Liras they had tobuy from their counterparts in option contracts.


7

The role of currency options in that particular crisis and further crises is clear. Thehedging strategy delayed the crisis and made it much more sudden and violent. If currencyoptions were not available, investors in Liras would have been obliged to hedge in futuresmarkets. Selling pressures would have been progressive, signalling to the central bank howsevere the speculation was. With the use of currency options, the dynamics in the foreignexchange market was different. Heavier sales of the attacked currency were triggered by theconcentration of option strike prices close to the upper margin of the band. In Italy, whenthe central bank raised its interest rate to defend its currency, it launched out an irresistibleoutflow of capital instead of the desired inflow. The market failure precipitated theexchange crisis. Under the magnitude of the speculative attack, the foreign exchangemarket became illiquid. The exchange rate stumbled much more than warranted onfundamental grounds.

Basically similar dynamics occurred in the early 1995 dollar crisis, but with thehelp of more sophisticated option contracts. Those exotic derivatives were "knock-out" calloptions of the dollar. The hedge funds which bought them forced the banks that sold themto set up complicated hedging positions. The hedging strategies exacerbated the marketturmoil as prices fell and those hedges had to be unwound.

What is amazing is that dollar markets themselves can become illiquid under stressgenerated by dynamic hedging strategies which oblige dealers to sell into falling marketsand buy into rallies, greatly magnifying price volatility. In early March 1995, the worseningof liquidity conditions in the New York FOREX market was signalled by widening spreadsbetween "bid" and "offered" prices in inter-dealer trades to three times their norms, even inthe dollar-DM market, the deepest in the world.

Knock-out options are particularly vicious financial innovations. They are calloptions on the dollar that give the right to buy dollars at a pre-set price. But instead ofalways expiring worthless when they are out-of--the-money, like standard options, they getknocked out when the dollar falls below a specified trigger level. Another type of thresholdis embodied in the markets. The dealers who sold these options against the yen accumulatedlarge dollar positions to hedge the risk that the yen could break below the knock-out priceof 95 to the dollar. When it happened dealers engaged in heavy trading, selling dollarsmassively to decumulate their positions as much as the options contracts got knocked out.

The latest dollar crisis showed on a large scale other destabilising features whichexacerbated the impact of the core self-reinforcing process due to the growth of optionsmarkets associated with dynamic hedging strategies. The combustible fuelling the processis borrowing. The leverage available to currency traders through bank credit lines amountsup to one hundred times their capital or more. Leverage gives them the momentum to bringon extra-volatility in their markets. And markets become thin endogeneously because of thecomplexity of the hedging strategies embodied into the derivatives and because currencytraders are trend-followers, launching collective moves of buying as markets rise and sellingas markets fall.

Liquidity problems in normally liquid markets can occur when they are hit byrecurring waves of huge selling orders in quick succession in the course of a single tradingday. It occurred several times in dollar markets in the hectic days of March 1995.


8

I.2. The 1994 worldwide bond market slump

After the tightening of US monetary policy in February 1994, the yield on longterm bonds increased two to three hundred basis points within three months. The rise wasunprecedented in scope, spreading all over the industrial countries. It induced huge losses.According to the 1995 BIS Report, capital losses reached $ 1500 billions (approximately10% of OCDE GDP), the heaviest in fifty years. Furthermore, the highly synchronisedincrease in yields was accompanied by a much higher volatility which was alsointernationally correlated. The latter phenomenon was all the more surprising than ithappened in the lowest interest rate countries with long-standing inflation performance, likeGermany and the Netherlands.

Several features were intriguing in this period. The overall slump in bond pricesoccurred while inflation was low and diminishing, economic prospects were promisingrobust growth. The move was synchronised between markets despite opposite monetarypolicies : the FED was tightening whereas European central banks were loosening, after theturmoil on European currency markets had abated. Foreign bond rates did apparentlyrespond to the Fed's action more intensely than US rates!

It is farfetched to invoke news on fundamentals to explain such a generalisedbehaviour of interest rates when economic conditions and prospects in the US, Europe andJapan pointed to different stages in the business cycle. Neither future inflation nor a realcommon shock is of any help. One has to resort to the assumption of a market failure. Aspeculation on a continuous decline in long-run US interest rates had gained momentum inlate 1993, as a way to correct the steep slope of the yield curve at the time. Like thecurrency markets, the bond market was grasped by mimetic behaviour. Huge positions weretaken on the bet of lower interest rates and financed by very high leverage, by means ofshort term borrowing and by the use of derivatives. Institutional investors, mutual fundsand banks, all relied on the same financing pattern. On the BIS opinion, the systematicrecourse to very high leverage is an indication of an aggressive attitude toward risk. It ismotivated by excess competition in financial markets where individual performance isjudged relatively to the average of the same category of financial institutions.

Competition was also the stimulus for international arbitrage. When US bond ratesdeclined in 1993, international investors sought profitable arbitrage in Europe, Canada andLatin America. Derivatives allowed them to hedge against currency risk. Because they arewell suited to separate interest and currency risks, derivatives permit direct interest arbitragewhile capping against the risk of currency volatility. It ensues that internationaldiversification of institutional portfolios becomes more sensitive to nominal interest ratedifferentials [Goodhart, 1994]. It amounts to an alleviation of preferred habitats andconversely a greater substitutability between bonds denominated in different currencies.But it means that investment strategies become less selective and more prone to changes inmarket sentiment. In the period under review the end-result was the building of portfolioswhich were strongly exposed to interest rate risk. Such portfolios are very sensitive tounexpected bad news, to which they overact. Indeed leverage, which was magnifyingcapital gains when yields were declining, was aggravating capital losses when theyupturned. Investors hurried altogether to close their positions in order to repay their shortterm loans or honour the margin calls they suffered with their losses on derivatives.


9

Special factors in the markets seemed to have fed on the destabilising momentumof the precipitous debt deflation. In the US, a study by the Federal Reserve Bank of NewYork has highlighted the role of the mortgage debt market. With the rise of long rateshouseholds slowed down the (anticipated) repayments on their mortgage debt. The outcomewas a longer duration of these securities. To restructure their portfolios, the institutionalinvestors which hold the bulk of the mortgage debt had to sell public bonds of equivalentduration, adding up to the overall selling pressure.

International investors who arbitrage several markets apply stop-loss orders andportfolio insurance strategies on interrelated markets. They launch together automatic saleson multiple positions to limit their losses when specific thresholds are reached. Since lossesin the US triggered sales of bonds in Europe and waves of sales fed on themselves with thefall in prices, the moves were highly correlated among markets.

As mentioned before, the fall in price went along with an unexpected and abruptrise in the volatility of bond yields with a high international correlation starting in mid-February. It happened as if traders had discounted specific information on particularmarkets. This peculiar behaviour may be assigned to liquidity problems encountered bycritical market participants which span several markets and which are highly dependent onleverage. With the downturn of the markets, capital losses became more likely and liquidityneeds more pressing. They made risk aversion change abruptly. Non-resident investorswithdrawing from local markets were particularly responsive. According to the BIS, thesales of securities by non-residents were positively correlated with the variation in implicitvolatility on the German bond market. It is like a massive shrinking in internationalfinancial integration. The temporary retreat on national preferred habitats, confirmed by theincrease in the share of domestic public bonds, was an important factor of the increase inboth historic and implicit volatilities.

The other type of highly leveraged market participants are market markers. Theirliquidity problems in period of stress can occur in usually liquid markets, be they currencyor interest rate markets. Because of their critical role in the smooth functioning of themarket, a retrenchment of key market participants can induce disorderly conditions whichincrease uncertainty for all market participants.

I.3. Futures markets : the demise of Barings and Metallgesellschaft

The Barings crisis was well publicised. The news burst out in late February 1995 :Barings had lost more than $ 1 billion on massive bets in exchange-traded Nikkei futures bya 28-year-old trader in its Singapore office. This event teaches a lot about the deficienciesof control systems in long-standing, well-established Houses. There was no separation ofresponsibilities between trading functions on one hand, margin payments and dailysettlements on the other hand, there was no detection of the excess of potential losses fromderivatives exposure over the entire capital of the firm by internal control at higher level ofmanagement. Neither there was any detection by host or home country regulators.

It is important to notice that the systemic implications occurred not with OTCderivatives but with exchange-traded futures which are generally considered as safe marketsthanks to their legal procedures, daily clearing and margin calls. In some respect it is true.When Baring's inability to pay its margin requirements was revealed on February 24, quickaction was taken according to the rules of the Exchanges. Barings' customer trading


10

accounts were separated from its proprietary accounts, the former being transferred toanother clearing firm, without customers incurring losses. Singapore Monetary Authoritiesannounced that they were standing behind the clearing house. However there was a lack ofinternational co-ordination.

The Bank of England, supervisor of the failing bank was muddling through to workout a market solution, essentially concerned with London's reputation. Her top officers'attitude contrasted markedly with Mary Shapiro's decisive action. The chairman of theCommodity Futures Trading Commission (CFTC), who had already the experience of theunwinding of the Drexel case, knew enough about the chain reaction potential of a massivefailure, endangering the Singapore International Monetary Exchange (SIMEX) itself.

When the investment bank failed on February 27, Singaporean regulators frozefutures trades through Barings and doubled the amount of margins or collateral required tohold the dealers' positions. That self-protecting move threatened to be counterproductive,because it raised fears among global dealers about SIMEX'S fragility. They suspected thatthe required higher collateral might be used to bail out the Exchange's own losses instead ofbacking up their own positions. Therefore several US dealers were reluctant to pay themargins. But refusal to pay the extra margin would have caused dealers to forfeit theirfutures positions. SIMEX would have probably collapsed with the most dire consequenceson the Japanese Stock Market when the banking problems were already worsening.Moreover the Exchange's collapse would have drawn some member firms along and spreada confidence crisis in all futures markets worldwide.

Aware of the potential consequences, the CFTC's chairman called urgentlySingapore's top regulators to issue a joint statement assuring that the additional marginswould not be used to cover Barings' losses. Ms Shapiro did also persuade Japaneseauthorities not to freeze permanently client money as a result of Barings' failure. Soquieted, dealers paid up their margin calls and increased their collateral.

This episode shows how systemic risk can stem from the linkages between futuresmarkets. The answer to this threat is a big improvement in the cooperation amongExchanges. The need is hampered by the competition for market share between them.However sharing data for the knowledge of large market positions taken on by dealers withmultiple positions on various Exchanges is crucial. This knowledge could permit marketregulators to cooperate in supervising individual dealers worldwide.

By contrast, The Metallgesellschaft case was self-contained. But it teaches anotherdisturbing lesson. The German system of corporate governance, so efficient within theregulated and intermediated German financial system, failed to monitoring the marketstrategy of a US subsidiary using futures markets.

Metallgesellschaft A.G. (MG) is the fourteenth largest industrial German firm. Inlate 1993 and early 1994 its US subsidiary (MG Corporation) reported staggering losses onits positions in energy futures and swaps, ultimately exceeding $ 1.3. billion. Only amassive $1.9 billion rescue operation by 150 German and international banks kept MGfrom going into bankruptcy.


11

During 1993, MG's US oil trading subsidiary, MG Refining and Marketing(MGRM) established very large derivatives positions in energy futures and swaps, fromwhich it would profit handsomely if energy prices were to rise. But energy prices fellsharply during the latter part of 1993, causing MGRM to incur unrealised losses and margincalls on these derivatives positions in excess of $ 900 millions.

MGRM's derivatives activities were part of a complex hedging strategy. Itsderivatives position was used to hedge price exposure on forward-supply contracts thatcommitted it to supplying approximately 160 million barrels of gasoline and heating oil toend-users over the next ten years at fixed prices.

MGRM hedged this price risk with energy futures and OTC swaps. However thereis nothing like a perfect hedge. Hedging mainly shifts the risk parameters and, if well-conceived, diminishes its magnitude. MGRM's hedging strategy was set up to protect theprofit margins in its forward delivery contracts by insulating them from increases in energyprices. MGRM hedged its forward-supply commitments of 160 million barrels with short-dated futures and swaps, barrel for barrel (a hedge ratio of one). It had to choose short-maturity contracts because the longer ones were not liquid enough to match itscommitments of such a high magnitude. But it had to roll forward continuously itsderivatives positions to maintain its hedge.

The in-built risk of this strategy resided in the roll over. It was profitable when theforward markets were in backwardation and loss-making when they were in contango. ThusMGRM's belief that its rollovers would be profitable amounted to a bet of a higherfrequency of backwardation than of contango. According to F.R. Edwards, the short-datedhedging strategy exposed MGRM to a rollover risk o the order of 15 percent of its price risk[Edwards and Canter, 1995].

MGRM's misfortune when nearly prices fell sharply in late 1993 highlights twofeatures that make derivatives risky even for hedging : the myopic view of marketparticipants and the illiquidity of the longer-dated segments in the markets.

In predicting future rollover returns, the firm used historical data ten years backand made two critical assumptions : first history would repeat itself, meaning that thestructure of energy futures markets would not change significantly in the future ; second ahistory of only ten years is long enough to uncover the long-run price relationshipsnecessary to infer long-run equilibrium rollover returns. Both assumptions are simply notwarranted. Large prediction errors result from too short time series : first from the shift instructural price relationships when fundamental economic conditions change irreversibly ;second from high variances in the distribution of rollover returns relative to mean rolloverreturns.

A less risky hedging strategy was not available at a reasonable cost. In principle,MGRM could have used strips of long-dated futures exactly matching the dates of itsforward delivery contracts. Had MGRM contracted with OTC dealers, the latter wouldhave had to use a similar rollover strategy to hedge their own long-term exposure toMGRM. They would have charged MGRM with a high risk premium to cover thelikelihood of an adverse price change and the possibility of MGRM's default which is quitehigh over a ten-year period. Further more those illiquid contracts must be terminated by


12

direct negotiation with the dealers, a serious drawback for MGRM exposed to the earlycash-out options embedded in its forward delivery contracts.


13

I.4. Emerging markets : the Mexican crisis

In the early 1990's, developing countries began taking bold actions of financialliberalisation. In undertaking this process, Mexico had become a highly-praised model forother so-called emerging market economies, particularly in Latin America. The response ofthe financial community was enthusiastic. Because of the financially-engineered recessionin the industrial countries, yield were declining there. As already mentioned before,institutional and other investors were lured by the mood of governments in favour ofopening their financial systems to international capital. Therefore the structure of financingwas dramatically altered both on the debtor and on the creditor sides in many developingcountries.

The inherited public debts were securitized by Brady bonds, which are dollar-denominated securities backed by US Treasury bonds. These instruments were welcome byhedge funds and investment banks, because they are traded in relatively thin markets whereit is easy to create volatility. One strategy applied to Mexican and Venezuelan bonds was tomake a collar by buying both call and put options on the same underlying bonds atpredetermined strike prices. Then the hedge funds began buying the underlying asset toincrease the volatility of the bonds, which in turn increased the volatility of the call options.

The banks who had sold the options would hurry buying the bonds themselves tocover their short positions. Once again the hedging strategies were magnifying priceincreases, attracting more conservative investors who were lured by the rising tendency ofthe markets. A diversification in local currency debt occurred also, thanks to the policies ofpegging the local currencies which allowed the same type of gambling as the convergencetrade in Europe. Governments were able to issue short-term debt, as did the Mexicangovernment with Tesobonos, which are dollar-linked securities. The local private sector,banks and non-banks alike, was able to rest on the renewed confidence in the liberaleconomic policies pursued by their government in order to issue large amounts of debtowed to non-resident investors.

The fundamental reasons for the Mexican crisis are well-known and do not need tobe restated in the present paper [Pisani and Sgard, 1995]. I am interested in the immediatecause of the crisis which was an acute liquidity shortage in early 1995 and, in theperspective of the present study, I am concerned with the systemic implications.

In late 1994 it appeared that the Mexican authorities would face an unmanageableliquidity problem in the first half of 1995. The amount of maturing debt owed to non-residents reached the dazzling figure of $ 60 billion, due to the careless short-termborrowing in 1994. Nonetheless the myopic view of the markets confirmed countless priorexperiences. The risk premium on Tesobonos over US Treasury bills was 2 percent on thewake of the peso dizzying decline in December 1994! It jumped after the crash to reach 20percent in February 1995 [Sachs, Tornell and Velaseo, 1995]. The same was true forpremia on Brady bonds and the forward exchange rate discount which did not anticipate thecrisis.


14

On the amount of debt held by foreigners and coming due, more than a third wereTesobonos, nearly a third were interbank lines of credit to Mexican commercial banks and afifth was owed by the private non-bank sector. Only $ 8.5 billion constituted the externalobligations of the Mexican foreign debt. On the creditor side, the overwhelming part (morethan 90 percent) was held by non-official investors. Moreover a projected deficit on currentaccount of $ 7 billion in the first half of the year should be added up to the $ 60 billion debtmaturing. Against this pending obligation, the available liquid funds were utterlyinadequate. Prospects for raising new money on international capital markets or rollingover part of the existing maturing debt had dried up. This is the permanent feature of afinancial crisis. Seemingly limitless market liquidity sources just weeks before simplydisappear when the strangled debtor needs them the most. When all official sources hadbeen collected, the financing gap facing Mexico was about $ 55 billion in the first half of1995, meaning a financial crisis of a huge magnitude. There was no question that systemicimplications would be dramatic and that the stability of the international financial systemwas a stake.

To reveal the systemic linkages, a method is to guess what could have happened ifthe international monetary authorities had let it burn out, i.e if they had sat and looked at theso-called "market solution". The financing gap would have been closed without any newprivate money inflow and no rollover of existing claims. The economic slump would havebeen much worse than even the contractionary adjustment really experienced by Mexico inthe first half of 1995. Because of the huge demand for dollars, the foreign exchange valueof the peso would have collapsed without anyone knowing which value would be a floorunder the dollar price of the peso. The peso value of the private sector debt would havereached extravagant amounts ; chains of bank and non-bank bankruptcies would havedevastated the domestic economy because those agents were unable to finance their externaldebt repayments.

In the midst of the macroeconomics upheaval, the Mexican government wouldhave had no other choice than making allowance for the market failure. It would haverepaid Tesobonos in pesos and unilaterally suspended the repayment on part of Mexico'sdebt. It would have been forced to freeze repayments of the debt owed by banks,effectively imposing stringent capital controls.

The most dire consequences would have ensued. In Mexico the stock market wouldhave collapsed and interest rates skyrocketed. The financing of the economy eitherdomestic or foreign would have receded to a low ebb. With the freeze of foreign depositsinto Mexican commercial banks, trade credit would have dried up. The market solutionwould have possibly resulted in its opposite : an end to the liberal economic policies.Considering the model role played by Mexico in the propaganda for liberal reforms indeveloping countries, some other emerging market economies would have surely reviewedtheir strategies. Indeed significant contagion effects would have been likely, at least inLatin America. More generally a global retrenchment of capital flows from emergingmarkets would have tightened balance of payment constraints and stifled world growth.


15

II. DERIVATIVES AND FINANCIAL FRAGILITY

Behind the diversity of the reviewed episodes of financial turmoil, one can detectrecurrent attributes of financial fragility. Those attributes are latent in the present structureof international markets and are activated in time of stress and deterioration of marketsentiment. Because the financial disturbances occurring in fragile structures have thepotential to spread over markets, the dynamic linkages involved in the propagation ofmarket risks are sources of systemic risk. Derivatives pertained to the relevant externalitiesin all the aforementioned crises.

In conformity with the findings, I will first draw lessons from the empirical cases todescribe the sources of systemic risk stemming from financial markets. Then I willinvestigate some characteristics of present day derivatives markets that are conducive toexternalities which can foster systemic risk : destabilising price dynamics in period of stress,uncertainty about credit risk assessment, vulnerability to liquidity risks due to theconcentration of market making in OTC derivatives.

II.1. Lessons from the episodes : the sources of systemic risk in financialmarkets.

In the theories of systemic risk, liquidity problems are rightly singled out as themost prominent factor which is capable of propagating local financial disturbances [Davis,1992]. For economic agents a shortfall of liquidity alters decisively the budget constraint.It entails forced sale of assets and makes it more difficult to roll over liabilities [Minsky,1982]. It induces profitable transactions to be deferred and denies expected cash flows toothers, giving rise to cumulative income contractions. In a debt market, liquidity stringencycan engineer a market failure in the following sense : market participants become quiteuncertain of the price level which will balance demand and supply. Being unable toanticipate a bottom under the price slump caused by the selling pressure, potential buyers ofthe depreciating asset which could supply the needed liquidity stay put. Therefore volatilitycan increase tremendously, losses get huge and participants rush to recoup their losses, thustransmitting the selling pressure to other markets.

In a narrow view of financial crises, it is asserted that there is a need for concernonly if banks are affected As the bulk of liquidity comes from bank deposits, onlydisturbances which can launch out sequential runs on deposits throughout the bankingsystem can also significantly affect aggregate liquidity. This viewpoint appears to seriouslyunderrate the way liquidity is generated in present-day wholesale debt markets.

Securities markets are increasingly relied on as repositories for liquidity. Liquidsecurities are substitutes for lower-yielding cash or demand deposits and a complement tomatching of liabilities by assets over the longer term. If banks are actively engaged in thesecurities business, they can undergo a liquidity crisis directly because they rely on therelevant market for funding or because they are unable to meet their commitments toprovide backup facilities to other market makers. This direct linkage is adding up to theindirect one referred to by monetarists, whereby suspected losses of banks fromunderwriting or market making lead to doubts on the part of depositors regarding theirsolvency. Moreover, if derivatives are involved in the process of financing asset holdings inthe underlying securities markets, their own market liquidity may supplement the moretraditional interbank market liquidity as the central feature of money markets.


16

Be that as it may, the dealers' role of banks in OTC derivatives has hugelyincreased since the late 1980's. OTC derivatives are conceived to be bridges betweensegments of securities markets that were not perfectly arbitraged beforehand. As ways ofgetting liquidity, they wipe out the specificity of the interbank market. As instruments forhedging risk, they erase the separation between national and international markets.

With the multiplication of bank failures in different countries, with the rise of non-performing assets and the inadequate amounts of provisions for losses, the creditworthinessof international banks deteriorated markedly from 1990 onwards. Counterparty risksbecoming more acute, banks reduced their participation to the interbank market. BISstatistics reflected the overall contraction of deposits in the international interbank market.A growing share of bank liquidity needs was covered by negotiable instruments provided byinstitutional investors. The latter were lured into the short-term debt market by the lack ofprofitable assets due to the recession. As it happened with earlier financial innovation, achange in the pattern of financing, which was the market response to peculiarcircumstances, has become a permanent feature of financial practices afterward because it isless costly than financing with interbank deposits in a period of declining interest rates.

When an institutional investor is bringing forward liquidity by buying short-termdebt instruments, it has to hedge against the interest rate risk (and the currency risk forinternational investors). OTC derivatives are particularly suitable to provide tailor-madeinstruments. A small cluster of large international banks stepped into the fast-developingOTC derivatives to catch the opportunity of dealers' profits in writing the customerproducts.

Therefore the wholesale market for liquidity has undergone a profound overhaul.In order to investigate if it is more or less prone to systemic risk than the traditionalinterbank market, one should know the answer to the crucial question : who are the ultimatewriters of derivatives (i.e options and swaps)? If, in a low interest environment, dealers canhedge in their own markets, buying contracts sold by the investing community itself, thenmarket risk is effectively diversified. No dynamic externality leading to systemic risk canoccur. However, it is extremely unlikely that such an optimal pattern of risk diversifyingcan hold in volatile environments or in uncertain macroeconomic situations, when thesentiment of investors is changing abruptly and collectively in response to shocks. Theepisodes examined earlier in the first part of this paper give credence to this assertion.Systemic risk can take root in such circumstances.

In so far as sources of systemic risk stemming from dynamic linkages betweenmarkets for liquidity are concerned, the following lessons can be drawn from the episodes.

i. Market illiquidity can force banks acting as market makers to rely on dynamichedging and to effectively convey the liquidity gap onto the underlying markets

Market makers are supposed to satisfy end-users (buyers and sellers of securitiesand derivatives) and to maintain an orderly market : i.e to limit uncertainty caused by pricefluctuations, so that a market's overreaction can be countered by fundamentals-based tradersacting on the basis of underlying economic considerations. In their capacity they areexposed to losses incurred if they keep standing against one-way selling. Because losses


17

might be too large with respect to capital or because credit might be too expensive or toorisky to finance the accumulation of depreciating assets, market makers can give upsustaining market prices at any predetermined level [Bingham, 1992]. The market failureentails liquidity to dry up in the particular market and uncertainty to raise substantially thevolatility of prices. Therefore dealers and a number of other market participants have toresort to dynamic hedging to hedge their positions. As observed in the episodes, marketsfor options are vulnerable to this self-reinforcing mechanism. If there is a concentration ofoption contracts at the same perceived critical price, dynamic hedging will exacerbate pricemovements on the underlying asset market (and feed back on the option price through therise in volatility) and disturb cash market liquidity [Brockmeijer Report, 1995].

As seen in the bond market slump, margin and collateral calls on derivatives arehigher in times of heightened volatility. To fulfil their commitments, market participantsfeel a pressure to liquidate their underlying assets, thus aggravating the price decline. Asfor the potential to create systemic risk from speculative behaviour by individual marketparticipants, the Barings crisis is a good example. Absent decisive actions by regulators,uncertainty could have induced dealers to forfeit their positions causing SIMEX to collapse; which would have provoked a slump in the Japanese Stock Exchange, thus worseninginsolvency problems of Japanese banks.

ii. Numerous reasons account for the possibility of one-way selling in present-daymarkets where competition is intense and price expectations are affected by complexparameters

As mentioned above, institutional investors play a major role in the wholesale debtmarkets and associated derivatives markets. They are in competition for market shares andthey are sensitive to short-term profits. It entails the same response to common signals, thesame portfolio insurance strategies and the direct influence of each other's behaviour.Under those conditions it can be shown that the market demand for risky assets can beunstable and the resulting market price highly volatile [Artus, 1995]. TheMetallgesellschaft case and the Venezuelan Brady Bonds are good examples of elusiveexpectations based on too short data series.

Leverage exacerbates the price instability induced by the lack of individualdiversity in the market structure, constraining investment strategies to being less selective.By facilitating market arbitrage, leverage and the related use of derivatives spread the extra-volatility from one market to another.

Asymmetric information is another type of market imperfection giving rise toinsiders' deliberate creation of volatility and to the occasional impediment of market makingin thin markets. When dealers face a group of market participants more informed than themon the parameters that move the market price, they need to charge a higher spread or restricttransactions to offset losses made on dealing with insiders [Leland, 1992]. This externalitycan lead to a market failure when spreads are widened and suppliers of liquiditydiscouraged with the tightening grasp of insider traders. If there are sizeable fixed costs inmarket making, the amount of liquidity is inversely responsive to the cost of transactions. Avicious circle can thus occur when insider trading widens the bid-ask spread, which in turnreduces the volume of trade and induces liquidity suppliers to withdraw.

II.2. Destabilizing price dynamics in period of stress: the role of options


18

Several episodes in the foreign exchange markets highlighted the role of optionsand option-like instruments in disturbing the dynamics of the underlying prices at times oflarge changes in those asset prices. The diverging dynamics rest upon positive feed-backswhen investors trade on the perceived expectations of other investors. Such effects are in-built in the dynamic hedging activities of the writers of options. It is not to say thatderivatives play a leading role in sharp market price movements. But they play a larger rolein the subsequent asset price changes (Hanoun Report, November 1994).

With dynamic hedging of options exposure, heavier trading occurs in theunderlying assets under a period of stress. Sharp price movements entail a change in thedelta of options which gives rise in turn to heavier purchases (sales) when prices rise (fall).

But there is more to say. The risk taken by option writers can be unintentionallyvery serious considering their management practices. This might be a transitory period,however, because the state of risk management is improving rapidly. Nonetheless, the waythe non-linearities of options values are dealt with in the overwhelming majority of riskmanagement systems is liable to very large errors for broad fluctuations in underlying assetprices.

J.P. Morgan's Risk Metrics offers the most standardized methodology for valuingmarket risk in whole portfolios embodying options or not. The methodology is based uponthe concept of value-at-risk.

This concept is appealing because it measures different risks in terms of a commonmetric: an amount of losses relative to a standard unit of likelihood. It is why value-at-riskcan aggregate risk across instruments, trading units and markets. More precisely, value-at-risk is an estimate of the potential changes in the value of a portfolio, based on a statisticalconfidence interval of changes in market prices, that are likely to occur some proportion ofthe time [Fisher Report, 1994]. Therefore value-at-risk incorporates two importantelements of risk: the sensitivity of a portfolio to changes in underlying prices; the volatilityof these underlying prices But value-at-risk is not a risk limit. It is a measure of the likelyexpected declines in portfolio value that will be exceeded some proportion of the time.

Operational for daily management, the concept has some drawbacks for prudentialpurposes. It is based on historical data of the relevant asset prices. They do not takeaccount of extreme market conditions (i.e. abrupt changes in price of a large magnitude)especially if the observation period is short, as the Metallgesellschaft case taught us. Themethodology assumes unlimited market liquidity, which is inappropriate under stressbecause liquidity problems are a major cause of brisk unanticipated departure of pricevolatility from an historical pattern. The concept relies on the tail of the probabilitydistribution of the stochastic variations in prices, under the assumption of normality of thedistribution If the assumption of normality is not warranted, large errors can be made if theactual distribution has fat tails. Finally, and it is the point with options, price variationsunder stress conditions can be on the order of several times the standard deviation. If thesensitivity of the instrument is assumed to be a linear function of the change of theunderlying market price, the linear approximation to compute conveniently the value-at-riskof the instrument can lead to huge errors in the likely loss under a given confidence intervalwhen the actual function is non-linear.


19

In dealing with option, the usual practice to compute changes in value is to use aTaylor series development of the second order (delta an gamma approximation) on theoption pricing formula, around the current market price of the underlying asset. Inneglecting higher order terms and the influence of the other variables (passage of time,changes in interest rate and in implied volatility), the variance of the option value can bemade a function of only delta, gamma and the variance of the underlying price [J.P.Morgan's Risk Metrics, 1995].

Serious inaccuracies may result from this practice when prices change widely fromthe current market price [Estrella, 1995]. Even higher order Taylor series do not warrant abetter approximation, because the Taylor series may not converge for large fluctuations inthe underlying asset price. The alternative is very cumbersome and needs sophisticatedsystems and a good deal of expertise. It requires using the exact option pricing formula tocompute variations in option value under a large set of movements in the underlying assetprice. This is the simulation approach. If the option pricing formula is reliable, the risk canbe more accurately measured on conditions that a large number of scenarios can begenerated. Because options are incorporated into aggregate portfolios, pricing modelsshould revalue each instrument under all possible scenarios.

II.3. Uncertainty about credit risk: the swap market.

Counterparty risk is a source of cash flow problems with long-term tailor-madeswaps, because banks are very active in this fast-developing segment of derivatives.Systemic risk can occur when a default by one intermediary is imposing liquidity constraintson counterparties not able in turn to meet their payment obligations. A defaulting swapdealer is led to early termination of his swaps. If market making for longer maturities ofOTC swaps is concentrated, the amount owed by the defaulting party can be a significantsource of funds for end-users who could have difficulties to fund the unexpected shortfall ofcash by borrowing. Therefore the interdealer market is the crucial segment. To hedge theirpositions in the swap market itself, dealers must own each other's portfolio thoughtransactions in the interdealer market. Concentration in this market makes it more difficultto achieve full hedging with two possible consequences. First, as far as dealers are banks,they offset the unhedged leg of their net swap position in the interbank market. Second, thefailure of one dealer entails the search for liquid funds by its counterparties, which ends upin borrowing from banks. Therefore, even if the liquidity-generating process has becomemore roundabout with the proliferation of negotiable debt-cum-derivatives instruments, it isstill based upon the interbank market. The latter is exposed to sudden unexpected demandfor liquid funds to close positions or satisfy margin calls triggered by shocks in derivativesor other financial markets. The interbank market is the lender-of-before-last-resort. Thereis no evidence, to say the least, that the securitization of liquid debt has relieved theinterbank market from its central role. This linkage shall be kept in mind while analyzingthe sources of risk in the swap market, which is the main type of derivatives for funding.

Credit risks are of critical importance in the swap market and are very difficult tomeasure. A counterparty A suffers a credit loss on a swap if his counterparty B defaultsand if the swaps has a positive value for A at the time of default. The magnitude of the lossis the difference between this positive value and the amount that can be recovered from thedefaulting counterparty. It depends on two stochastic processes: the probability ofcounterparty default on one hand, the potential credit exposure on the swap portfoliobetween A and B on the other. The latter is highly volatile (especially with currency swaps)


20

because it depends on the change in the values of the marked-to-market instruments whichstructure the swaps. The most insuperable difficulty to reach a correct estimate of creditlosses is the fact that the correlation between the stochastic processes governing defaultprobabilities and financial market variables is unknown. Ignoring the joint stochasticbehavior of defaults and credit exposure can lead to greatly underestimate credit risk[Duffee, 1994]. However there is no way to integrating market and credit risk underexisting valuation models and information systems. Nevertheless we know by historicalexperience of business cycles and financial crises that this correlation exists. Indeed, thiscorrelation is the gist of systemic risk. It is not significant in normal conditions but it risesdramatically when the financial system becomes fragile in a period of heightened volatilityof asset prices and high leverage. A sharp decline in asset prices tends to concur with anincrease in the probability of default of non-financial and financial firms.

There are fundamental reasons for systemic risk not to be detected by sophisticatedstatistical analysis. To estimate potential credit exposure on a swap contract, it is necessaryto approximate the probability of future changes in credit exposure (the stochastic variable)over a relevant time interval. The estimate of the probability distribution is drawn from asimulation (calibrated with historical data) of a large number of randomly-generated timepaths for the underlying financial variables. The resulting numerical distribution of futurecredit exposure is used to calculate the expected potential credit exposure or the"maximum" potential credit exposure associated with a confidence interval.

This structured Monte Carlo approach presupposes that future data are generatedby the same process that generated historical data. Therefore the only source of uncertaintyappears implicitly to be the future underlying prices, not the model generating the paths. Itis tantamount to say that the stochastic process which moves financial prices is stationarythrough time. That reference does not hold under the conditions of stress which pertain to aregime change. In regime changes, stochastic processes are decisively altered. When afinancial crisis occur, volatility cannot be inferred from the time series included in thedataset. It increases widely with the sharp decline in the level of prices. A regime changecan only be detected by experience. Recognizing such extreme market conditions permitsto draw information not included in standard simulations, but which is the raw material ofstress tests. However there is no reliable technique to discriminate between a regimechange and a sequence of events in the tail of the historical distribution using only a smallnumber of observations [Kupiec, 1995]. A large number of observations are necessary toseparate acceptance or rejection of a regime change. But the financial crisis will be wellunder way and losses severe indeed! It is why experience cannot be replaced by statisticalanalysis. But for experience to be helpful, it needs a strong involvement of highermanagement and a control system which imposes severe limits to trading desks. Recentepisodes show that financial institutions are very far from being aware of the pitfallsembodied in their most sophisticated methods.

Acknowledging that valuation models of potential credit exposure and credit lossesare flawed under extreme conditions does not preclude from accepting their conclusionsabout the parameters of credit risk on a swap portfolio under more normal conditions.

The current credit exposure of a single contract for a counterparty is equal to zeroif the value of the contract is negative (out-of-the-money contract) or to the value contract if


21

it is positive (in-the-money contract). The credit exposure of the portfolio is the sum of thevalues of all the in-the-money contracts. Potential future credit exposure on a given timeinterval is a multiple (depending on the confidence interval) of the standard deviation of thechange in credit exposure of the portfolio. The analytic determination of this variable ispossible under the assumption that all individual swaps contracts have the same standarddeviation of their change in value over the time interval and that their values do not changesign in the interval. Under these assumptions, the standard deviation of the change in creditexposure of the portfolio is an increasing function of three variables: the standard deviationof the change in value of the individual contract; the number of in-the-money contracts inthe portfolio; the average correlation of changes across all in-the-money contracts[Hendricks, 1993]. The interpretation of the first two variables is self-evident. The thirdone is the extent to which changes in the value of the in-the-money contracts move together.It will rise whenever interest rates move in the same direction. Since the averagecorrelation is extremely volatile with interest rate changes (and exchange rate changes forcurrency swaps), potential credit exposure can change rapidly as well. It is why hedging inthe swap market itself is difficult. It is especially the case for dealers who have positivelycorrelated contracts with multiple counterparties. Even if the average correlation is weaklypositive, the dealer's potential exposure increases with the number of contracts.

II.4. Concentration of market making in OTC derivatives.

Potential future credit exposure in swap markets and option risks are very sensitiveto changes in volatility of underlying asset prices. They are major sources of risk for marketmaking firms. When unable to hedge in their own market or when trapped by insufficientliquidity, those firms transmit the imbalances to other markets and to third parties, thuseffectively being vectors of systemic risk. The reactions of critical market makers fosteringpositive feedbacks can occur as responses to changes in market sentiment due to shocksentailing uncertainty in credit risk assessment and (or) extreme market price fluctuations.An hedging overhang in OTC markets can spread over derivatives Exchanges or theinterbank market for liquidity needs. Serving as hedging or lending of before-last-resortfacilities, these markets can be strained under one-way conditions in the original OTCderivatives.

The likelihood of one-way conditions depend in turn on the breadth and depth ofmarket making in the original markets. Many reasons explain why both characters whichmake for the dissemination of risk are not always present in OTC markets. Riskmanagement is complex and loaded with fixed costs. They are not only investment costs ofdeveloping and maintaining large information and data processing systems. Even moreimportant are learning costs to cope with ever-changing instruments which preclude thetraditional routines of risk assessment. It requires building powerful statistical models formarking-to-market widely diversified portfolios and for estimating risk profiles of largevectors of correlated instruments.


22

The cost structure is conducive to increasing returns to scale. It is why dealers are asmall number of large banks which straddle a range of interrelated OTC derivatives. Optionwriters face financial institutions who are likely to behave homogeneously under a commonprice shock, for instance in the currency markets. Swaps dealers are likely to becounterparties of end-users who want to hedge interest rate risk. As we have seen above,potential credit exposure depends on the mixture of pay-fixed and pay-floating swapsdealers have in portfolios as a result of meeting their customers' demands. Swaps of liketype tend to be highly positively correlated; swap of opposite type tend to be highlynegatively correlated. In situations when end-users fear the same move of interest rates,those who rush for cover are likely to hedged against the same direction of interest ratechange. In that case dealers will have a positively correlated portfolio of swaps with end-users. If they are concentrated, it is not clear that they can all hedge their positions bycontracting negatively correlated swaps with one another. It is why large potential creditexposure can be transmitted to other markets.

The vulnerability to spillover effects increases with the swap duration (theprobability of counterparty default gets higher), the volatility of underlying prices and thesensitivity of the instrument to the magnitude of future price changes (potential credit riskexposure gets higher), the number of counterparties for each dealer (correlation betweenlike type swaps gets higher) the vulnerability decreases with the amount of capital provisionper dealer and with the number of co-dealers in the market.


23

CONCLUSION

The picture presented hereabove looks gloomy. But it is not the attribute ofintrinsic flaws in derivatives markets, nor is it without remedy. The destabilizing processes,acknowledged in reviewing recent episodes of financial turmoil in the first part of the paperand analyzed further in the second part, occur under abnormal circumstances. More thanonce the extent of disorderly conditions is contained in narrow segments of financialsystems. Dynamic externalities do occur. But imbalances are absorbed when they reachmarkets which are deep and liquid enough.

However the recurring pattern of central bank interventions and large scalerestructuring of whole financial sectors with the backing of public money in a wide array ofcountries sounds as a warning. Systemic risk is a latent feature of present-day financialsystems. There is a legitimate concern for unlikely but highly damaging market dynamics,when individual behavior worsens overall financial conditions.

It is encouraging that a debate between market participants and regulators is beingheld with the objective of promoting a new approach to risk prevention. An interactiveprocess is set up whereby relevant information can be improved thanks to better methods ofcontrolling market risk at the firm's level and to more comprehensive disclosure requiredfrom all major market participants. But risk prevention is not the end of the story.Whatever its enhancement, unanticipated shocks will still happen in particular marketscausing a rush for exit. To avert contagion into interrelated markets, there is a case formore stringent capital standards against market risk and for the adoption in OTC markets ofthe very safety rules already introduced in Exchange markets. The G10 Committee ofcentral banks is addressing those issues and making proposals on both informationdisclosure and more demanding standards.

i) Disclosure of market and credit risks

The pricing of market risk is not always accurate, especially as complex portfoliosare structured with interrelated instruments. Market risk should measure the potential forlosses on the global trading portfolio, encompassing both derivatives and on-balance-sheetsecurities. As soon as September 1994, The Fisher Report acknowledged that internal riskmanagement practices improved within the firms but public disclosure lagged for behind. Agap has widened between the ability of financial firms to manage their own risks and theirinability to assess the riskiness of other participants.

Under stress, a low market transparency is conducive to the one-way sellingpressure highlighted in the present paper. If information is lacking about the risk exposureof market makers, a general suspicion about their financial situation can easily spread out.Funding difficulties encountered by one of those firms appear similar to outsiders for othermarket makers, if available information is too poor to permit screening. Rumors of aliquidity squeeze take over which can be self-fulfilling.


24

A meaningful disclosure should apply to all major market participants, besystematic and structured in a common framework to be understood by outsiders. The roleof regulators is crucial to implement the procedure, because individual firms will not goahead unilaterally. Indeed, limited information entails that a firm revealing moreinformation than others about its risks may reasonably fear that outsiders perceive it to beworse off than market makers which abstain form disclosing. On the contrary, if regulatoryauthorities announce a set of quantitative and qualitative standards to be met by majormarket participants for the sake of market transparency, outsiders will see that the faster afirm adjusts to the required disclosure, the safer its management practices and the better itscontrol system. A positive dynamic process can be initiated whereby relevant informationdisclosure is leading to enhanced disclosure practices.

The BIS has stated the principles of disclosure based upon internal models of riskassessment by financial firms. The formal framework is based upon the value-at-riskconcept, which gives information to market participants on the propensity of a financial firmto take market risk. All market makers over a conventional threshold measured by the sizeof trading account assets should be subjected to disclosure on the same grounds.

But revealing daily value-at-risk is far from being sufficient. The holding periodmust be long enough to guard against the risk of being locked into unbearable positions. Asdepicted in the second part of the paper, this can be the case with the non-linear pricebehavior of options. The BIS considers that ten days is a minimal holding period for thenon-linear properties of options being captured. The assessment of risk on theseinstruments should be made by using the full pricing formula and the effect of the change inthe valuation of option contracts on the whole portfolio should be displayed.

As far as credit risks are concerned, the exposure of firms can extend well into thefuture in the case of swaps. At a minimum they should disclose the distribution of thereplacement costs of their current outstanding positions by counterparty rating. But currentexposure does not give a complete picture of credit risk. A more accurate measure ispotential future credit exposure which shows how the exposure to credit risk can vary as theresult of changes in market prices. A precise estimate of potential future credit exposure israrely made by banks, let alone its disclosure. There is a large room for improvement inthis respect. But more elaborate assessments require sophisticated simulation models.

These models should also be developed to provide regulators with more relevantindicators of the vulnerability of market markers to systemic risk, namely by stress tests.For the time being stress tests are still in infancy, largely ad hoc and run with widelydifferent methods from one firm to another. According to the BIS, promoting a routine andrigorous program of stress testing would be a very valuable tool for market participants andregulators. It would enable a better understanding of the dynamics of systemic risk andwould be the primary commodity for the definition of indicators of the vulnerability ofwhole markets to this most extreme form of failure. To carry out this particularinformation, stress tests should simulate market losses and changes in credit risk exposureas consequences of the following events : large and positively correlated price moves in keyrisk factors, worst case combination of macroeconomic shocks, major changes ofcorrelation between risk factors, loss of market liquidity, failure of an importantcounterparty.


25

Because markets are highly interlinked and because market makers straddle a lot ofmarkets, coordination between market regulators and bank supervisors should become anessential feature of a comprehensive prudential policy at the international level. TheBarings crisis displayed blatant deficiencies in this respect. Coordination is a necessarycomplement of better disclosure to exploit the information revealed by stress tests for thesake of systemic risk prevention.

ii) More demanding standards

The debate between central banks of the G10 and banks acting as dealers insecurities and derivatives about the proper capital charge to be held against market risk islively to say the least. The authorities have accepted the principle of building theirsupervisory framework on the internal measurement of market risk by market participants.But setting up the prudential rules which make the principle operational and still preserverobust financial structures is a tricky job.

In the course of the paper, many reasons were given why the daily value-at-risk isnot the appropriate measure of capital requirement. Moreover the present state of the art inindividual banks differs widely, so that they can come up with a broad range of value-at-riskmeasures for the same portfolio under the same present market conditions. It is why aminimal normalization in the methodology, a routine disclosure with a larger array of riskmeasures, an involvement of higher management in tighter internal control systems, are allnecessary conditions for a direct use of internal measures.

Absent these conditions, the supervisors propose to base the capital charge on thevalue-at-risk taken on a 10-business day holding period and apply a conventional multiplierof 5. The proposal is hotly contested by the banks who claim that it will lead to a much toohigh and too expensive capital requirement.

Only an improvement in risk measurement and disclosure will help reduce the gapbetween the conflicting views of both parties. The regulators should have enoughinformation to detect if the value-at-risk reported by individual banks is particularly low andwhy it is so. In particular, because correlations between risk factors are held constantbetween two revision dates and change abruptly on revision dates, the daily value-at-riskmay understate the change in the underlying volatility, then jumps artificially on a specificrevision date.

As supervisors want traders to protect for longer risk, they demand a 10-businessday holding period to report value-at-risk. To account for the volatility of the measure itself,the supervisors can make an averaging of the preceding reports over a period of severalmonths backward or take the medium of the range defined by the x% worst figures.

The scaling-up factor is introduced to make for all the risk factors which areunderestimated or completely ignored by the value-at-risk methodology. They are thefactors conducive to systemic risk. As long as those factors are not well known or are notquantified even if understood, the multiplier will be conventional and vulnerable to critics.Only an accumulation of knowledge drawn from long series of stress tests could give anorder of magnitude for the multiplier, which should ideally measure how much themagnitude of risk can increase when market conditions shift from normal to abnormalcircumstances.


26

iii) Organizing OTC derivatives markets

The Barings' failure which occurred on Exchange-traded contracts plainly showedthat the unwinding of the positions of an insolvent entity could be efficient and fast. Therewas a safety net with multiple lines of defense : customer trading accounts and proprietaryaccounts were separated on SIMEX, the Singapore Monetary Authorities were standingbehind the SIMEX clearing house to guarantee settlements, the doubling of marginrequirements on the Nikkei 225 futures contract brought on liquidity after the CFTCintervened decisively.

This safety net is completely lacking on OTC contracts. A market maker's defaultis much more time-consuming to unwind. Potential systemic implications are much largerwhen positions are not daily marked-to-market according to market prices, when variationmargins are not called on a permanent basis with changes in potential losses, when noclearinghouse mechanism can take over the customer accounts of a failing firm and transferthem to another clearing firm.

To become robust to liquidity problems induced by heavy losses of importantdealers, OTC markets should be made more alike Exchange-traded markets. This meansintroducing multilateral netting mechanisms and collaterals sensitive to changes in value-at-value risk on market exposures and to changes in potential future credit risk exposure.Therefore the requirement of collaterals is closely dependent on progress in the disclosureof potential losses. Multilateral netting meets formidable obstacles to be implemented : thewide variety of OTC contracts, the legal claims to the trading assets of the failing firm undernational bankruptcy laws, the lack of transparent market prices. Therefore multilateralnetting is dependent on a rule-setting process acquiring legal status and harmonizing theworking of OTC markets under a coordinated supervision.


27

REFERENCES

AGLIETTA M. et MOUTOT P., 1993, Le risque de système et sa prévention, CahiersEconomiques et Monétaires de la Banque de France, n° 41.

AGLIETTA M. et de BOISSIEU C., 1994, Les marchés dérivés de gré à gré et le risquesystémique, COB.

ARTUS P., 1995, Anomalies sur les marchés financiers, Economica.

BINGHAM T., 1992, Securities markets and banking: some regulatory issues, in H.Cavanna ed., Financial innovation, Routledge, London.

BIS 65 th Report, 1995, Basel, June.

COMMITTEE of the Central bank of the GIO:FISHER Report, 1994, Public disclosure of market and credit risks by financial

intermediaries, BIS, September.

HANNOUN Report, 1994, Macroeconomic and monetary policy issues raised by thegrowth of derivatives markets, BIS, November.

BROCKMEIJER Report, 1995, Issues of measurement related to market size andmacroprudential risks in derivatives markets, BIS, February.

DAVIS E.P., 1992, Debt, financial fragility and systemic risk, Clarendon Press, Oxford.

DAVIS E.P., 1994, Market liquidity risk, in D. Fair and R. Raymond eds., Thecompetitiveness of financial institutions and centres in Europe, Kluwer AcademicPublishers.

DUFFEE G., 1994, on measuring credit risks of derivatives instruments, Federal ReserveBoard, September.

EDWARS F. and CANTER M., 1995, the collapse of Metallgesellschaft: unhedgeablerisks, poor hedging strategy or just bad luck?, Columbia University and AmericanEnterprise Institute, March.

ESTRELLA A., HENDRICKS D., KHAMBHU J., SHIN S., WALTER S., 1994, Theprice risk of options positions, Federal Reserve Bank of New York, Quarterly Review,Vol. 19, n°2, summer-fall.

ESTRELLA A., 1995, Taylor, Black and Scholes: series approximations and riskmanagement pitfalls, Federal Reserve Bank of New York Research Paper, N°9501.

HENDRICKS D., 1993, Netting agreements and potential credit exposure, Federal ReserveBank of New York.


28

KINDLEBERGER C.P., 1978, Manias, panics and crashes, Basic Books.

KUPIEC P., 1995, Techniques for verifying the accuracy of risk measurement models,Federal Reserve Board, April.

MISHKIN F., 1994, Preventing financial crises: an international perspective, TheManchester School Supplement, Blackwell Publishers.

MINSKY H., 1982, The financial instability hypothess, capitalist processes and thebehaviour of the economy, in Kindleberger and Laffargue eds, Financial crises: theoryand policy, Cambridge University Press.

PISANI-FERRY J. et SGARD J., 1995, Leçons mexicaines, La lettre du CEPII, n° 134,avril.

REMOLONA E., 1993, The recent growth of financial derivative markets, FederalReserve Bank of New York Quarterly Review, vol. 17, n° 4, winter

RISKMETRICS, 1995, J.P. Morgan, 3rd ed, May.

SACHS J., TORNELLA A, VELASCO A., 1995, The collapse of the Mexican peso: whathave we learned?, NBER Working Papers, n°5142, June.

SCHWARTZ A., 1992, Real and pseudo financial crises, in M. Bordo ed, Financial crises,vol I, chap 1, Cambridge University Press.


29

LIST OF WORKING PAPERS RELEASED BY CEPII2

1995

"Why NAFTA might be Discriminatory", Lionel Fontagné, document de travail n° 95-12,décembre

"Régionalisation et échanges de biens intermédiaires", Lionel Fontagné, MichaelFreudenberg et Deniz Ünal-Kesenci, document de travail n° 95-11, décembre.

"The Geography of Multi-speed Europe", Philippe Martin et Gianmarco I.P Ottaviono,document de travail n° 95-10, novembre.

"The Political Economy of French Policy and the Transmission to EMU", Christian deBoissieu et Jean Pisani-Ferry, document de travail n° 95-09, octobre.

"L'importance des exclus de l'intégration monétaire en Europe", Philippe Martin, documentde travail n° 95-08, novembre.

"Asymétries financières en Europe et transmission de la politique monétaire", VirginieCoudert et Benoit Mojon, document de travail n° 95-07, septembre.

"La mesure du capital éducatif", Pierre Villa, document de travail n° 95-06, septembre.

"Capital humain, mobilité des capitaux et commerce international", Pierre Villa, documentde travail n° 95-05, juin.

"L'Europe à géométrie variable : une analyse économique", Jean Pisani-Ferry, document detravail n° 95-04, avril.

"Comparaison de l’efficacité énergétique des pays d’Europe centrale et orientale avec celledes pays de l’OCDE", Nina Kounetzoff, document de travail n° 95-03, mars.

"L'organisation de la politique économique dans un cadre stratégique", Pierre Villa,document de travail n° 95-02, mars.

"Interest Rates, Banking, Spreads and Credit Supply: The Real Effects", Fernando Barran,Virginie Coudert, Benoît Mojon, document de travail n° 95-01, mars.

1994

"L'après-CAEM : La dynamique des échanges entre les pays de Visegrad", DominiquePianelli, document de travail n° 94-16, décembre.

"CEEC Export to the EC from 1988 to 1993: Country Differentiation and CommodityDiversification", Françoise Lemoine, document de travail n° 94-15, décembre. 2 Working papers are circulated free of charge as far as stocks are available; thank you to send yourrequest to CEPII, Sylvie Hurion, 9, rue Georges Pitard, 75015 Paris, or by fax (33.1.53.68.55.03)


30

"Union monétaire et convergence : qu'avons nous appris ?", Jean Pisani-Ferry, document detravail n° 94-14, décembre.

"Chômage et salaire en France sur longue période", Pierre Villa, Document de travailn° 94-13, novembre.

"Croissance et spécialisation", Frédéric Busson et Pierre Villa, Document de travail n° 94-12, novembre.

"The International Monetary System in Search of New Principales", Michel Aglietta,doucument de travail n° 94-11, septembre.

"French and German Productivity Levels in Manufacturing: a Comparison Based on theIndustry of Origin Method", Deniz Unal-Kesenci et Michael Freudenberg, document detravail n° 94-10, septembre.

"La réunification allemande du point de vue de la politique économique", Agnès Bénassy,Pierre Villa, document de travail n° 94-09, septembre.

"Commerce international, emploi et salaires", Olivier Cortes et Sébastien Jean, document detravail n° 94-08, août.

"La fonction de consommation sur longue période en France", Pierre Villa, document detravail n° 94-07, juillet.

"Réglementation et prise de risque des intermédiaires financiers : la crise des prix d'actifs audébut des années 1990", Benoit Mojon, document de travail n° 94-06, juillet.

"Turquie : d'une stabilisation à l'autre" Isabelle Bensidoun, document de travail n° 94-05,juillet.

"Economic Policy Strategies to Fight Mass Unemployment in Europe: An Appraisal.",Henri Delessy et Henri Sterdyniak, document de travail n° 94-04, juillet.

"Transmission de la politique monétaire et crédit bancaire, une application à cinq pays del'OCDE", Fernando Barran, Virginie Coudert et Benoît Mojon, document de travail n° 94-03, juin.

"Indépendance de la banque centrale et politique budgétaire", Agnès Bénassy et JeanPisani-Ferry, document de travail n° 94-02, juin.

"Les systèmes de paiements dans l'intégration européenne", Michel Aglietta, document detravail n°94-01, mai.

1993

"Crises et cycles financiers : une approche comparative", Michel Aglietta, document detravail n°93-05, octobre.


31

"Regional and World-Wide Dimensions of Globalization", Michel Fouquin, document detravail n° 93-04, septembre.

"Règle, discrétion et régime de change en Europe", Pierre Villa, document de travail n° 93-03, août.

"Crédit et dynamiques économiques", Michel Aglietta, Virginie Coudert, et Benoît Mojon,document de travail n° 93-02, mai.

"Les implications extérieures de l'UEM", Agnès Bénassy, Alexander Italianer et JeanPisani-Ferry, document de travail n° 93-01, avril.

1992

"Pouvoir d'achat du franc et restructuration industrielle de la France 1960-1991", GérardLafay, document de travail n° 92-04, décembre.

"Le Franc : de l'instrument de croissance à la recherche de l'ancrage nominal", MichelAglietta, document de travail n° 92-03, décembre.

"Comportement bancaire et risque de système", Michel Aglietta, document de travail n° 92-02, mai.

"Dynamiques macroéconomiques des économies du sud : une maquette représentative",Isabelle Bensidoun, Véronique Kessler, document de travail n° 92-01, mars.

financial market failures and systemic riskfinancial market failures and systemic risk 3 resume dans...

Documents